hmaraniam

Zero-dependency language identification for Hmar.

"Hmar a ni am?" ("Is it Hmar?")

hmaraniam is a Python library that identifies Hmar text and tells it apart from English and the related Kuki-Chin / Zo languages (Mizo, Paite, Thadou, Vaiphei, Gangte, Zou).

Maintained by the Hmar Heritage Foundation as part of the Hmar Heritage Archival Project.

Features

• Frequency-weighted detection: Scores each token using corpus log-frequencies (log(1 + count)) built from 2 Hmar Bibles and 583 verified web articles, covering 45,042 unigrams. Core structural words like chu, chun, and an score higher than rare or loanword tokens.
• Dual diacritic scoring: Reports casual_hmar_ratio (ASCII-normalized for QWERTY typing) and formal_hmar_ratio (exact diacritic matches).
• Sibling Zo language resolution: Separates Mizo, Paite, Thadou, Gangte, Zou, and Vaiphei from Hmar using dialect-exclusive particles and per-language vocabulary lists.
• Separate confidence scores: hmar_confidence answers "how Hmar is this text?" independently of detected_language_confidence, which rates the overall classification call.
• Consistent output shape: Every call returns the same dictionary structure, including word counts, frequency-weighted ratios, sibling scores, and diacritic breakdowns.
• Raw text cleanup: Strips HTML tags, Markdown syntax (bold, italic, links, code blocks), URLs, and email addresses from pasted input before scoring.
• Custom vocabulary: Pass custom unigram sets, extra domain words, or custom stopword lists directly to the Detector.
• Offline first: Ships with a bundled shard so it works without a network call. CDN sync via jsDelivr is available when you need the latest data.
• Zero dependencies: Pure Python standard library. No PyTorch, TensorFlow, NumPy, or spaCy.

Design

hmaraniam answers one question: "Is this text Hmar?" It does not correct spelling or modify the input.

Diacritic normalization. Mobile keyboards produce inconsistent accent codepoints. casual_hmar_ratio strips diacritics before matching so ṭha and tha both score against the same vocabulary entry.

Token boundaries. When you need precise control over how a token is defined (for example, whether mithiem-hai counts as one word or two), pass a pre-tokenized list. The library will not re-split it. For plain strings, it tokenizes by word boundary.

Vocabulary, not grammar. The score reflects dictionary overlap, not sentence structure. A list of valid Hmar words scores the same as a grammatical sentence with the same words.

Installation

pip install hmaraniam

Output schema

{
  "language": "hmar",
  "hmar_confidence": 0.9842,
  "detected_language_confidence": 0.9842,
  "sibling_heuristic": false,
  "mode": "basic",
  "scores": {
    "casual_hmar_ratio": 0.9524,
    "weighted_hmar_ratio": 0.9103,
    "formal_hmar_ratio": 0.8095,
    "english_stopword_ratio": 0.0000,
    "sibling_zo_stopword_ratio": 0.0000,
    "hmar_stopword_ratio": 0.1429,
    "unknown_words_ratio": 0.0476,
    "total_words": 21,
    "hmar_words_count": 20,
    "non_hmar_words_count": 1,
    "unknown_words_count": 1,
    "english_stopwords_count": 0,
    "sibling_zo_stopwords_count": 0,
    "hmar_stopwords_count": 3,
    "sibling_lang_scores": {},
    "hmar_diacritic_words_count": 17,
    "non_hmar_diacritic_words_count": 0,
    "total_diacritic_words_count": 17
  }
}

Usage

Quick start

import hmaraniam

# Authentic text quote from L. Keivom archive (Coleman Factor, 2002)
sample_text = "Khawvel fe dan phung ei en chun, ram le hnam damna thuruk chu lien lema intel le insung khawm, zai khat le trong khata luong khawm a nih."

result = hmaraniam.detect(sample_text)
print(result)

Pre-tokenized inputs

If you need to define token boundaries yourself (for example, to treat mithiem-hai as a single token rather than two words), pass a list, JSON file, CSV, or line-delimited TXT. The library scores each entry as-is without re-splitting.

Supported formats

1. JSON array (tokens.json):

   [
     "khawvel",
     "fe",
     "dan",
     "mithiem-hai",
     "pathien",
     "hnenah"
   ]
   

   Usage: hmaraniam.detect("tokens.json") or CLI hmaraniam tokens.json

2. CSV (tokens.csv):

   token
   khawvel
   fe
   dan
   mithiem-hai
   pathien
   hnenah
   

   Usage: hmaraniam.detect("tokens.csv") or CLI hmaraniam tokens.csv

3. Line-delimited TXT (tokens.txt, 1 word per line):

   khawvel
   fe
   dan
   mithiem-hai
   pathien
   hnenah
   

   Usage: hmaraniam.detect("tokens.txt") or CLI hmaraniam tokens.txt

4. Python list:

   tokens = ["mithiem-hai", "pathien", "hnenah", "khawvel"]
   result = hmaraniam.detect(tokens)
   

Raw text

Pass a plain string or a .txt file path and the library tokenizes it automatically. HTML, Markdown, and URLs are stripped before scoring.

# Raw text string
result = hmaraniam.detect("Khawvel fe dan phung ei en chun, ram le hnam damna thuruk...")

# Raw text file
result = hmaraniam.detect("path/to/article.txt")

Custom unigrams and stopwords

from hmaraniam import Detector

# Provide custom unigrams or extra domain vocabulary
detector = Detector(
    mode="basic",
    extra_unigrams=["customworda", "customwordb"],
    custom_stopwords=["and", "the", "with"],
    disable_default_stopwords=False
)

result = detector.detect("Khawvel fe dan phung...")

Modes

In v0.2+, the entire 45,042 frequency-weighted vocabulary is bundled directly into the library, operating 100% offline with zero network latency.

from hmaraniam import Detector

# Default / Basic mode (45k unified unigrams, offline)
detector = Detector()

# High mode (retained for backward compatibility, uses unified dataset)
high_detector = Detector(mode="high")

# Explicit offline flag
offline_detector = Detector(offline_only=True)

Datasets

• Hmar Unigrams (unigrams): 58,983 verified Hmar surface words and active loanwords.
• Corpus Archive (corpus-archive): Archival text corpus of Hmar literature and lexicons.

Error handling

hmaraniam raises standard Python exceptions:

import hmaraniam

# Raises ValueError for unsupported modes
try:
    hmaraniam.detect("Text", mode="ultra")
except ValueError as e:
    print(e)

# Raises TypeError for non-string input
try:
    hmaraniam.detect(12345)
except TypeError as e:
    print(e)

License

MIT License. Published by the Hmar Heritage Foundation.

Citation

@software{hmaraniam_2026,
  author       = {Hmar Heritage Foundation},
  title        = {hmaraniam: Zero-dependency language identification library for Hmar},
  year         = {2026},
  publisher    = {PyPI / GitHub},
  iso_code     = {hmr},
  glottolog    = {hmar1241},
  clade        = {Zo Languages},
  howpublished = {\url{https://github.com/hmar-heritage-org/hmaraniam}}
}